QMUL, School of Electronic Engineering and Computer Science
Centre for Digital Music Seminar Series
Kaustuv Ganguli (Post-Doctoral Associate, NYU Abu Dhabi)
Date/time: Wednesday 27th of November, 15h00-16h00
Location: Graduate Centre - Room 204
Number 18 on Campus map: https://www.qmul.ac.uk/media/qmul/docs/about/Mile-End_map-April2019.pdf
Open to students, staff, alumni, public; all welcome. Admission is FREE, no pre-booking required.
Title: How “different” is different: A perspective on melodic similarity in Indian art music
Abstract: Indian art music is predominantly an oral tradition with pedagogy involving the oral transmission of raga lessons. There exist text resources for musicology, wherein the lexicon finds its place in a rather prescriptive manner. Raga performance allows for considerable flexibility in interpretation of the raga grammar in order to incorporate elements of creativity via improvisation. It is therefore of interest to understand how the musical concepts are manifested in performance, and how the artiste improvises i.e. uses the stock musicological knowledge in “new” ways, while carefully maintaining the distinctiveness of each raga in the ears of trained listeners. Alongside addressing the issue of subjectivity, scalability, and reproducibility in musicological research, this work proposes novel methods for relevant music information retrieval (MIR) applications like rich transcription, melody segmentation, motif spotting, raga recognition. While a general engineering approach is to optimize certain evaluation metrics, we aimed to ensure that our findings are informed by musicological knowledge and human judgment. To achieve this, our approach is two-fold: computational modeling of the melody on a sizable, representative corpus; then validating the models through behavioral experiments towards understanding the learned schema by trained musicians. We propose computational representations that robustly capture the particular melodic features of the raga under study while being sensitive enough to the differences between ragas tested within a sizable, representative music corpus. To make a good foundation for tuning of hyper-parameters, we exploit the notion of “allied ragas” that use the same tonal material but differ in their order, hierarchy, and phraseology. Our results show that computational representations of distributional and structural information in the melody, combined with suitable distance measures give insights about how the aspect of raga distinctiveness is manifested in practice over different time scales by creative performers. Finally, motivated by the parallels between musical structure and prosodic structure in speech, we present listening experiments that explore musicians’ perception of ecologically valid synthesized variants of a raga-characteristic phrase. Our findings suggest that trained musicians clearly demonstrate elements of categorical perception in the context of the technical boundary of a raga.
Bio: Kaustuv Kanti Ganguli is a professional vocalist, an engineer, and a budding musicologist trained in Hindustani music tradition. After finishing his PhD at the Indian Institute of Technology Bombay, India, Kaustuv has joined the New York University Abu Dhabi, UAE, as a Postdoctoral Associate. He has got several invited talks, international journal and conference publications to his credit. His thesis work includes research on artificial intelligence in music, with a focus on computational musicology. Currently he is working on generative models and human-computer interactions towards developing interactive visualisations for Emirati music. Kaustuv has been performing in various cities of India as well as abroad and is a recipient of several prestigious accolades. His vision is to promote experimental works on music production, analysis, and pedagogy with a view to converging his knowledge in music and technology fronts. Recently Kaustuv has taken up his passion for perfumes to study a possible synesthetic mapping with musical sounds.